Parallel Algorithms for Limited Views THz Tomography

Jose Antonio Cantoral Ceballos and Krikor B. Ozanyan

School of Electrical and Electronic Engineering, The University of Manchester, Manchester M13 9PL, United Kingdom

ABSTRACT

Tomography imaging of industrial subjects requires compliance with a particular environment. This results in severe limitations on the access and the deployable hardware resources, defining the situation of imaging by systems with limited resources, to be addressed by the future generations of imagers. This means that faster and more economic implementations of data acquisition and image reconstruction are called for, enabled by the latest software and hardware developments and system design.

This contribution offers a review of the approaches and methods for implementation of parallel and distributed computing for the needs of Tomography, particularly when the efficient use of dedicated resources is a priority, e.g in the case of embedded Tomography systems. We envisage generic hardware which can be configured as a multiprocessor environment, programmable to implement data acquisition, reconstruction and visualisation. The state of the art is examined and issues are highlighted from a computer science perspective, as well as from the point of view of current progress in available hardware and recent trends in reconstruction methods amenable to parallelisation.

Particular emphasis is given to THz Tomography from limited views and the potential of the recently introduced method of sinogram restoration by sinusoidal Hough transform (HT), which has been demonstrated up­to­date only as a sequential algorithm, but is amenable to parallel computation. Possible strategies for its deployment on distributed unutilized resources in a multi­FPGA system are also discussed, including fast centre­of­mass detection, without solving the standard inverse Radon transform.

Keywords Parallel computing, Hough Transform, limited views THz Tomography